Anthropic announced that its AI model Claude will incorporate invisible watermarks into generated texts to help with identification. However, the company noted the signals cannot determine the exact level of AI intervention or authorship.